Image Classification on MNIST non-i.i.d. (test)
93.08AccuracyFedMPDD
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FedMPDDBytes Budget=200000000, Target Acc=60, m=9002025.12 | 93.08 | 90,720,000 | — | |
| FedMPDDBytes Budget=200000000, Target Acc=60, m=15002025.12 | 80.67 | 151,200,000 | — | |
| FedMPDDBytes Budget=90000000, Target Acc=60, Model Architecture=LeNet, m=4002025.12 | 76 | 51,840,000 | — | |
| FedMPDDBytes Budget=90000000, Target Acc=60, Model Architecture=LeNet, m=8002025.12 | 57.8 | 93,440,000 | — | |
| QSGDBytes Budget=200000000, Target Acc=60, precision=8-bit2025.12 | 21.84 | 1,030,776,000 | — | |
| QSGDBytes Budget=90000000, Target Acc=60, Model Architecture=LeNet, Quantization=8-bit2025.12 | 18.66 | 359,924,000 | — | |
| FedSGD + LaplaceBytes Budget=90000000, Target Acc=60, Model Architecture=LeNet, Noise=Laplace (var=0.5)2025.12 | 12.75 | 1,482,230,400 | — | |
| FedSGDBytes Budget=200000000, Target Acc=602025.12 | 11.4 | 3,696,243,200 | — | |
| FedSGDBytes Budget=90000000, Target Acc=60, Model Architecture=LeNet2025.12 | 11.06 | 1,460,748,800 | — | |
| FedSGD + LaplaceBytes Budget=200000000, Target Acc=60, variance=0.0012025.12 | 11.04 | 3,696,243,200 | — |